This. I haven't seen meaningful lifts from changing button colors, wording or all these things I read on guru blogs. Most (all) of the tests I've seen reported on blogs fail miserably at the statistical test, not to mention no DoE, no blocking, no theory and a thousands other things that made the test invalid.

The only way I can see that a change in colors or wording would provide a lift is if the original was confusing, which shouldn't be the case.